CDs Not Lossy?

I lost my car keys this morning. That's "lossy", right?

 

For a COMPLETELY DIFFERENT REASON.

Anyway, since it ISN'T the same concept why give it the same name? One is a question of resolution, one is a question of compression. Two different things.

 

Yes, lossless only in the idealized case.
The point is that the terms lossy and lossless were already in use before computer science co-opted them to describe data compression.

 

No worries, I had my flame proof vest on.

I know what people mean by "lossy" - it does have a well accepted meaning that has nothing to do with downsampling - but there was something I found appealing about your thinking-outside-the-box interpretation of the term, so I couldn't help but want to defend it, even if the anal-retentive, see-the-world-in-black-and-white 5 year old in all of us reflexively rebels at such definitional corruptions. No better place to see that in action than an audiophile board.

 

The issue is with the term 'lossy'. It is a convenient shorthand for perceptually encoded. If you substitute perceptually encoded for 'lossy', you will understand that CD is not 'lossy' in the way that the term is used in data compression. Due to the negative connotations of the term 'lossy', coding schemes that merely try to improve the efficiency of storing the data became 'lossless'. A perfect perceptual codec (at a high enough encoding rate) would be indistinguishable from the master file from which it is created after decoding, since everything which has been discarded is inaudible. The issue with 'lossy' files is further signal processing, which introduces degradation, and the fact that no perceptual codec is ever perfect. Of course, trying to compress a file to its smallest absolute size also affects sound quality, hence the difference between 96 kb/s MP3 and 320+ kb/s MP3, say.

Sample rate conversion from a higher rate to a lower rate does lose information, but it is known that that is inherent to the process (we also know that sample rate conversion from a lower rate to a higher one adds information, but never the exact information that was lost). There is a world of difference in the two processes (sample rate conversion versus perceptual coding). Just because higher resolutions came along MANY years after CD was developed, does not make CD 'lossy' in any sense. It is a fixed container, and the only way to put a quart in a pint pot is to employ a lossless compression technique, which violates the CD standard.

 

Nobody said that these terms were invented or introduced by computer scientists. The only thing they did was to give them a specific meaning in the context of an specific field - just like other scientists are doing in other fields.

 

If anything it was "inside the box" and "black and white" thinking: "these two things are similar so let's call them the same thing even though they are completely different concepts"

 

Interesting.

 

If you can find them later in original shape, it was lossless.

 

Not really. Loss of resolution looses some content below about -90dB when recording at 16 bits. Everything above that level is "theoretically" preserved. OTOH, with lossy compression, lots of information (which the compressions algorithm has decided is not necessary) is deliberately thrown-out all together, and the signal levels for this information are WAY above -90dB.

If you measure the output from a 16 bit file, it will have the same spectral content as a 24 bit file. The only measurable difference becomes information which is below the LSB. But even a great deal of this information can be preserved below the theoretical recording limit of the system by the use of proper dithering.

OTOH, with a lossy compression algorithm, all sorts of information is deliberately thrown-out forever, and the signal will be measurably different. And the more complex the signal, the more significant the measured differences will be.

 

Exactly! Things, concepts, even people evolve. As our technology changes and grows and morphs along (high concept ideas here people) I think that theoretically one day the "set in stone definition" that some here are bashing others over the head with about a 16 bit 44.1 khz wav file is a baseline of being considered "lossless" will change. Think about it: why is 16/44.1 a standard that we all go by? Because someone decided that was good enough (and it was hard at the time to deal with much more data). But now, as higher and higher resolutions become more normal and embraced by *some* of the public, it *could* be nearing the time to raise the bar and declare that 16/44.1 isn't a good baseline anymore (one day!) and they will then be considered lossy...

If humans evolved to become much taller in 1000 years, then, yes, a six foot tall person WOULD be considered a midget.

The cold hard facts aren't set in stone, and even if they were, stones can be broken too...

 

...but we haven't developed the hearing of bats just yet. 16/44.1 is there for very good reasons. It was no accident.

 

Well, not that we've necessarily evolved super-human hearing in 1000 years or nothing, but, what about the idea of higher resolution formats having sympathetic frequencies that do affect the frequencies that we CAN hear? Overtones that we cannot hear that set up harmonic partials that DO colour the sound we hear. If those are sliced off like they are in a 44.1 khz recording you don't get any overtones or harmonic partials and a subtle difference in the tonal output happens. No, we don't have hearing of bats, but those unheard frequencies ride along with the ones we do hear and make the sound "more real" for want of a better term...

Just because you can't hear something doesn't mean you're not experiencing it!
http://en.wikipedia.org/wiki/Sympathetic_resonance
http://en.wikipedia.org/wiki/Overtone
While those links don't quite cover what I'm talking about, imagine something like this: you throw a rock in a pool and it makes a wave that travels at a certain speed and size and then you throw a bigger rock at an angle in the pool and it makes a bigger wave that travels faster... When the bigger wave catches up to the smaller wave they sort of join together and create a new and different larger wave. The super-high-harmonic frequencies we can't hear alter the ones we do hear.... At least that's my understanding of it... I'm no scientist but I don't see why simply stopping at CD resolution is a good thing!

 

https://people.xiph.org/~xiphmont/demo/neil-young.html

 

Good grief. You're regurgitating that again?
How many threads are you going to post that in?

 

This is my first time. Why? You've read it already? Good, then this discussion is moot

 

Mastering quality is more important than resolution.

 

Sorry, that link has been posted in a bunch of threads recently due to the Pono talk. That link gets to be old news real quick. There's already a dedicated thread for discussing Monty's opinions. No need to have it dumped in so many threads.

 

You're completely missing the point of this whole discussion. The OP's original question was:

"If they're often nowadays coming from higher resolution masters, why do we continue referring to CDs as lossless and not lossy?"

The answer: Because the distinction between lossy and lossless only makes sense in the context of compression (and as a CD contains uncompressed data, it qualifies as being lossless). Consequently these two terms are used in this context only. You're ALWAYS losing something when downsampling, just like you're ALWAYS losing something when initially recording a performance (depending on your equipment, positions of microphones, etc.), so it simply makes no sense to distinguish between lossy and lossless recordings or between lossy and lossless downsampling. As some kind of loss is always inherent to these processes, you simply omit the word "lossy" in this context, because it carries zero additional information, but creates confusion by immediately evoking the question if there are lossless variants of these processes (which there aren't). It's like going into a shop and asking for a "round CD with a hole in the middle". You wouldn't do this. You would simply ask for a CD. (Otherwise the salesperson's first question would likely be "Are there also non-round CDs? Or CDs without a hole?"). It's all about language and semantics.

Note that this has nothing to do with 16bit/44.1kHz. The exact same question (with the exact same answer) can be asked for a 24bit/96kHz SACD coming from an even higher resolution source.

 

Have you even bothered to read the thread? It has already been pointed out several times that WAV is NOT lossless. WAV is a container for PCM at some combination of word-length and sample rate. It is UNCOMPRESSED. CD is another container for PCM, and it too is UNCOMPRESSED, so is neither lossy nor lossless in the data compression sense. FLAC is lossless!

The fact that higher sampling rates/longer word-lengths emerged AFTER CD had been released does not change matters. 16/44.1 is considered a baseline because of the ubiquity of the CD format. It is the quality level most consumers know/can relate to. SACD/DVD-A/Blu-Ray/hi-rez downloads offering higher resolution do not make CD or 16/44.1 WAV/AIFF lossy.

No! The term lossy has a specific meaning, it does not refer to any digital signal processing step that may incur loss/add distortion, which would be every one!

 

"Lossless" and "Lossy" are terms to describe two different classifications of data compression.

 

They kind of are. That's what makes them facts.

 

This thread is priceless. Seriously.

 

Lossless is not lossless, a fact's not a fact ... what comes next? Priceless, indeed.

 

A few people have said this throughout the thread, but this strikes me as yet another example of faulty, narrow thinking. When you make a recording - at whatever fidelity, using whatever equipment, under whatever conditions - you are not "losing something". You are always gaining something. Unless the tape breaks, or the hard drive crashes, you're now in possession of something you didn't have before. It's never a loss.

From there, you can tamper all you want, but even then, those incremental steps towards a final master shouldn't be thought of as "losses". If the artist throws a ton of echo on the vocal track, that's not a "loss"; that's a gain towards his or her final vision of the record. After it's mastered, we can begin talking about losses - beginning immediately with downsampling to 44.1/16.

I'm not sure why people are still arguing; the OP has clearly indicated that the thread has served his purposes, and that he understands the issues now. A gort should close shop.